cloudbase-agent-ts

Use @cloudbase/agent-server's run() function with a LanggraphAgent wrapping your compiled StateGraph workflow. The workflow must be compiled with a checkpointer such as MemorySaver for development or CloudBaseSaver for production persistence.

Use @cloudbase/agent-ui-miniprogram's createAGUIBehavior mixin with a CloudbaseTransport configured with your bot ID. The behavior exposes this.agui.sendMessage() for sending messages and this.data.agui.uiMessages for rendering the conversation.

The LangGraph adapter wraps a compiled StateGraph workflow and provides ClientStateAnnotation with pre-wired messages and client tools fields. The LangChain adapter wraps agents created with LangChain's createAgent() and uses clientTools() middleware for client tool support.

Client tools are defined by the client and sent with the request. The server binds them to the model but routes client tool calls to END instead of executing them, letting the client run the handler locally and send the result back in a new request.

Yes, use CloudBaseSaver from @cloudbase/agent-adapter-langgraph, which stores checkpoints in Tencent CloudBase document database. Pass a userId extracted from the JWT Authorization header for multi-tenant isolation.

Do not use it for simple AI model calls without agent capabilities, CloudBase cloud functions without agent features, or general CloudRun backend services. It is specifically for AG-UI protocol agent deployments in TypeScript projects.